Why Waymo Paused Its Robotaxi Service in Multiple Cities

In May 2026, Waymo pulled its self-driving taxis off freeways in San Francisco, Los Angeles, Phoenix, and Miami. The company also shut down service entirely in Atlanta and San Antonio, Texas. The reason: the robotaxis struggled to handle two specific situations — construction zones and flooded streets — that caught the system off guard.

These problems may sound narrow, but they expose a real challenge facing self-driving car companies as they try to expand service. Testing a system in controlled conditions is different from running it in the messier real world.

What Went Wrong on Construction Zones

On a freeway, temporary construction is common. Orange cones redirect traffic. Signs tell drivers where to go. Workers stand in the road and wave cars through. For a self-driving car, all of this is noise compared to the permanent lane markings and traffic lights it learned from during testing.

Waymo's system relies on precise maps built beforehand, showing where lanes are and where traffic signals sit. When a construction crew rearranges the whole highway, those maps become outdated. The car has to figure out where to drive by reading temporary signs and spotting where other cars are going — abilities that the system has not perfected yet.

Flooding Caused Bigger Problems

The flooding incidents were worse. In Atlanta, a Waymo robotaxi got stuck in a flooded street for about an hour. The system could not tell how deep the water was or whether the road was safe to cross.

Flooding creates several overlapping problems for an autonomous vehicle. The computer vision system — the cameras that "see" the world — struggles to measure water depth. Lidar sensors, which work like sonar to map the environment, can bounce off water in confusing ways. And the routing system, which plans the best path to get somewhere, does not update fast enough to avoid hazards that change by the hour.

TechCrunch reported that Waymo issued a software recall — a remote update pushed to all its cars — to teach them to avoid flooded areas in San Antonio after similar problems.

Why This Matters at Scale

Waymo has announced a goal: one million paid rides per week by the end of 2026. These suspensions hit right as the company was pushing toward that target. That creates a real tension: how do you grow fast while also fixing problems that only show up when you try to grow.

This situation echoes patterns we saw before with other technologies. When GPS first came to phones in the early 2000s, it worked fine in open areas but confused itself in dense cities and tunnels. The difference then was that a wrong turn meant a few extra minutes driving. For a self-driving car on a freeway, safety matters more. Mistakes have higher stakes.

In this author's view, and based on watching similar transitions over three decades of covering technology, companies often underestimate how many unexpected situations will emerge once a system leaves the test track and meets real conditions at large scale. A few hundred test runs will not surface what happens when thousands of cars are on the road simultaneously, in different weather, in cities the system has never seen before.

The Technical Fix Path

Waymo's response shows a company with mature procedures for handling failures. When something goes wrong, it can stop operations in that area, figure out what broke, and push a fix to every car in the fleet without requiring a garage visit. That is a significant advantage over traditional cars.

The construction zone problem is solvable. The company needs better computer vision — systems that can read temporary signs and understand where workers are pointing. The flooding problem requires combining camera feeds, lidar data, and real-time weather information in smarter ways.

Whether these problems are just particular wrinkles or signs of deeper limits in how self-driving cars work is the bigger question. How Waymo answers that will shape not just when it hits one million rides per week, but when the rest of the industry reaches that milestone too.

For now, these suspensions show both the progress and the distance. Waymo can operate thousands of rides in some cities, but it cannot yet handle every situation it encounters. Closing the gap between those two facts is the work ahead.